Soleri
ENUAIT

Executable standards for agentic systems

From zero to a learning system in five minutes

Your first agent starts smart. It only gets smarter.

Three steps. No configuration files to write. No API keys required.

1

Install

One global npm package. That's it.

$ npm install -g soleri
2

Create your agent

The guided wizard walks you through each step — pick an archetype, name your agent, select domains, principles, skills, and tone. Most steps are just arrow keys and Enter.

$ soleri create ◆ What kind of agent? Security Auditor ◆ Display name: Sentinel ◆ Agent ID: sentinel ◆ Domains: security, code-review ◆ Principles: Security first, Zero trust... ◆ Tone: Precise ◆ Skills: 5 core + 5 selected ◆ Greeting: Auto ✓ Scaffolded 16 files, 2 facades, 66 ops ✓ Built npm install + npm run build ✓ MCP registered ~/.claude.json ✓ 10 skills installed TDD, debugging, planning, vault Restart Claude Code, then say "Hello, Sentinel!"
3

Activate and grow

Say hello in Claude Code. Your agent activates, captures knowledge as you work, and gets smarter over time. Add domains, install knowledge packs, or let the brain learn from your sessions.

# In Claude Code — activate the persona > Hello, My Agent! ✓ Persona activated My Agent — Frontend architecture advisor ✓ Vault ready empty — will grow as you work ✓ Brain tracking enabled — captures patterns from sessions # Add more domains any time $ soleri add-domain performance ✓ Created domain performance ✓ Rebuilt agent # Check everything is healthy $ soleri doctor ✓ Node.js v22.x ✓ Agent built, registered ✓ Hook packs a11y, clean-commits, typescript-safety

Wizard Reference

Everything the wizard offers at each step — so you know what you're choosing before you choose it.

1 Archetypes
2 Domains
3 Principles
4 Skills
5 Communication Tones
Step 1

Archetypes

Pick an archetype to pre-fill role, domains, principles, skills, and greeting. Or choose "Create Custom" for full control with guided examples.

Code Reviewer Catches bugs, enforces patterns, reviews PRs before merge
Security Auditor OWASP Top 10, dependency scanning, secrets detection
API Architect REST/GraphQL design, contract validation, versioning
Test Engineer Test generation, coverage analysis, TDD workflow
DevOps Pilot CI/CD pipelines, infrastructure, deployment automation
Database Architect Schema design, migrations, query optimization
Full-Stack Assistant General-purpose dev helper across the entire stack

Picking an archetype pre-fills domains, principles, skills, and tone. Choose "Create Custom" to set each one yourself.

Step 2

Domains

Each domain gives your agent a dedicated area of expertise with its own knowledge facade.

security Vulnerability scanning, threat modeling, secrets detection
code-review Pattern enforcement, anti-pattern detection, PR review
testing Test generation, coverage analysis, mutation testing
api-design REST/GraphQL contracts, versioning, error handling
performance Budgets, profiling, bundle size, query optimization
accessibility WCAG compliance, screen readers, keyboard navigation
architecture System design, boundaries, dependency management
database Schema design, migrations, indexing, query tuning
documentation API docs, READMEs, changelogs, code comments
devops CI/CD pipelines, infrastructure as code, deployment
Step 3

Principles

Guiding rules your agent follows when making decisions. Organized by category.

Quality
Simplicity over cleverness
Convention over configuration
Test everything that can break
Respect existing patterns
Safety
Security first
Fail closed, not open
Zero trust by default
Least privilege always
Developer Experience
Actionable feedback only
Explain the why, not just the what
Every comment includes a fix suggestion
Design for the consumer, not the implementer
Reliability
Graceful degradation over hard failures
Automate everything repeatable
Observability built in from day one
Every migration must be reversible
Step 4

Skills

Skills are capabilities your agent can use during sessions. Core skills are always included; optional skills are selected per archetype or manually.

writing-plans Structured multi-step planning before code changes
executing-plans Execute approved plans with review checkpoints
vault-navigator Deep-dive vault search and exploration
vault-capture Persist lessons learned to the knowledge vault
knowledge-harvest Extract patterns from completed work
brain-debrief Post-task intelligence summary and debriefing
code-patrol Scan for anti-patterns and code violations
test-driven-development TDD workflow: red, green, refactor
fix-and-learn Fix bugs and capture the lesson for next time
retrospective End-of-session retrospective and reflection
second-opinion Get a fresh perspective on tough decisions
onboard-me Guided codebase onboarding for new team members

Always included: brainstorming, systematic-debugging, verification-before-completion, health-check, context-resume

Step 5

Communication Tones

Sets how your agent communicates — from terse and factual to educational and explanatory.

Precise Direct, factual, minimal commentary
Mentor Educational, explains the "why" behind suggestions
Pragmatic Balanced, focuses on actionable outcomes

Keep going

How it works Vault, brain, and transport architecture explained. Your Agent Create, configure, and grow your agent's knowledge. Teams & Ops Shared vaults, knowledge packs, and Telegram bot.

Start building.

Open source. Apache 2.0.

Explore on GitHub Get started